DE — heise online

heise+ | How Google Protects Its AI Agents from Prompt Injections – An Interview

AIOctober 23, 20261 min read

This is a paywalled interview from the German tech site heise. It is about how Google protects its AI agents from prompt injections, which are tricks that fool AI into doing things it should not do. Only a short preview of the article was available.

What happened

heise talked with Viviana Sutedjo in Munich. She is a software engineer for cybersecurity and works on AI security at Google. The talk is about how Google keeps its AI agents safe from prompt injections. The free preview ends before the answers, so the full details are behind a heise+ subscription.

What is a prompt injection?

A prompt injection is a way to trick an AI model with clever wording. Early tricks were simple. People wrote their requests in another language, or told the AI a sad story to get past its limits. The name comes from SQL injection, a known computer security flaw. Today the tricks are harder. For example, a hidden calendar invite in an email can fool an AI agent that summarizes mail.

Why it matters

AI agents do real tasks, like reading email or using apps. If someone can trick an agent, they may leak data or make it act for them. So protecting these agents is an important job for companies that build them.

What to watch

Read the full interview with a heise+ subscription to learn how Google defends its agents.
Watch for new kinds of prompt injection attacks as AI agents get more power.
Look for new safety methods from Google and other AI companies.

Glossary

heise+: A paid subscription from the German tech news site heise that unlocks full articles.
Google: A large technology company that makes search, email, cloud services and AI tools.

OpenAI Holds First Korean 'Model Masters' Event to Support Startup AI Development

OpenAI held its first 'OpenAI Model Masters' event in Korea on the 23rd at its Seoul office, where 11 startups demonstrated services built on its models. The program, which provides API credits to startups recommended by venture capitalists, aims to help turn AI ideas into real products. OpenAI highlighted the selected startup Cutback's video-editing work as a standout example.

AIITmediaOct 26

Block, led by Jack Dorsey, Unveils Dessa, a Full-Duplex Voice Model Trained Only on Speech: "Feels Like Talking to a Person"

Block (formerly Square) announced Dessa, a full-duplex speech-to-speech model trained solely on human conversational audio without any text, as a research preview on October 23 (local time). Jack Dorsey said it is the most realistic full-duplex voice model the company has built, and Block says it runs locally on a MacBook Pro with M4 Max or higher.

AITechCrunchOct 26

Grok releases the first seven minutes of its AI-generated ‘Odyssey’

Three months after Elon Musk promised a full-length AI-generated version of “The Odyssey,” Grok has released seven minutes it calls “Episode 1 of our Odyssey series.” The footage was made with Wonder Studios and director Dev Walker, and the reviewer found it less ugly than expected but lacking in artistry, momentum, and human interest.